Re: [PATCH] arm64: contpte: fix set_access_flags() no-op check for SMMU/ATS faults

On Thu, Mar 05, 2026 at 05:33:25PM +0000, Catalin Marinas wrote:

Looking at the patch again, some more comments.

On Mon, Mar 02, 2026 at 10:37:51PM -0800, Piotr Jaroszynski wrote:

quoted

diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
index bcac4f55f9c1..9868bfe4607c 100644
--- a/arch/arm64/mm/contpte.c
+++ b/arch/arm64/mm/contpte.c

@@ -390,6 +390,23 @@ void contpte_clear_young_dirty_ptes(struct vm_area_struct *vma,
 }
 EXPORT_SYMBOL_GPL(contpte_clear_young_dirty_ptes);
 
+static bool contpte_all_subptes_match_access_flags(pte_t *ptep, pte_t entry)

More of a nitpick: since this checks both the flags and write
permission, I'd rename to something else. Maybe contpte_ptep_same() to
somewhat resemble pte_same() used by __ptep_set_access_flags().

pte_same() also compares the PFN though. I picked the _access_flags
suffix to match the ptep_set_access_flags() naming. I do agree the
aliasing of AF and "access flags" is unfortunate, but it seems
preexisting. I am updating the comments to be clearer in v2, let me know
if that works.

quoted

+{
+	pte_t *cont_ptep = contpte_align_down(ptep);
+	const pteval_t access_mask = PTE_RDONLY | PTE_AF | PTE_WRITE | PTE_DIRTY;

We can drop the PTE_DIRTY from the mask as it's not relevant to the
hardware permission. It probably doesn't matter in practice.

I think it's good to be consistent and just update everything while we
are doing it such that the sub-PTEs are in sync.

quoted

+	pteval_t entry_access = pte_val(entry) & access_mask;
+	int i;
+
+	for (i = 0; i < CONT_PTES; i++) {
+		pteval_t pte_access = pte_val(__ptep_get(cont_ptep + i)) & access_mask;
+
+		if (pte_access != entry_access)
+			return false;
+	}
+
+	return true;
+}
+
 int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
 					unsigned long addr, pte_t *ptep,
 					pte_t entry, int dirty)

@@ -399,13 +416,35 @@ int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
 	int i;
 
 	/*
-	 * Gather the access/dirty bits for the contiguous range. If nothing has
-	 * changed, its a noop.
+	 * Check whether all sub-PTEs in the CONT block already have the
+	 * requested access flags, using raw per-PTE values rather than the
+	 * gathered ptep_get() view.

It's not just about the access flag but AF, dirty and write permission,
all can be changed by this function (and only to a more permissive
setting).

Thanks, updating the wording in v2.

quoted

+	 *
+	 * ptep_get() gathers AF/dirty state across the whole CONT block,
+	 * which is correct for CPU TLB semantics: with FEAT_HAFDBS the
+	 * hardware may set AF/dirty on any sub-PTE and the CPU TLB treats
+	 * the gathered result as authoritative for the entire range. But an
+	 * SMMU without HTTU (or with HA/HD disabled in CD.TCR) evaluates

Or CPU equally, we don't force all CPUs in a system to support DBM.

Thanks, updating the wording in v2.

quoted

+	 * each descriptor individually and will keep faulting on the target
+	 * sub-PTE if its flags haven't actually been updated. Gathering can
+	 * therefore cause false no-ops when only a sibling has been updated:
+	 *  - write faults: target still has PTE_RDONLY (needs PTE_RDONLY cleared)
+	 *  - read faults:  target still lacks PTE_AF
+	 *
+	 * Per Arm ARM (DDI 0487) D8.7.1, any sub-PTE in a CONT range may
+	 * become the effective cached translation, so all entries must have
+	 * consistent attributes. Check the full CONT block before returning
+	 * no-op, and when any sub-PTE mismatches, proceed to update the whole
+	 * range.
 	 */
-	orig_pte = pte_mknoncont(ptep_get(ptep));
-	if (pte_val(orig_pte) == pte_val(entry))
+	if (contpte_all_subptes_match_access_flags(ptep, entry))
 		return 0;
 
+	/*
+	 * Use raw target pte (not gathered) for write-bit unfold decision.
+	 */
+	orig_pte = pte_mknoncont(__ptep_get(ptep));

This is fine since all should have the same PTE_WRITE bit.

Anyway, nothing major, so:

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>

Thanks!